In Switzerland, the second leading cause of death is cancer. In 2021, approximately 49000 new cancer cases were recorded ¹. One possible treatment option is proton therapy (PT). The unique physical properties of protons, particularly the Bragg Peak, enable highly localised energy deposition, reducing side effects and the integral dose overall and improving outcomes for patients with tumours in sensitive areas ².

However, despite its clinical advantages, PT remains largely inaccessible to many regions, particularly economically weaker countries, due to the cost of building and maintaining a PT centre. A portion of these costs arises from using large, complex gantry systems, which are required to rotate the proton beam around the patient and target the tumour from different angles.

Another challenge within PT is motion management. The patient's breathing motion, especially, reduces the accuracy of the treatment delivery within mobile tumours. This work explores the idea of magnet-free proton therapy delivery systems for mobile tumours, which would reduce the cost of the treatment and augment the possible treatment sites.

We want to create an artificial intelligence that can control therapy delivery and actively adapt to the patient's breathing motion using reinforcement learning (RL).